GorgiasB Argumentation in Practice
Argumentation in practice for real-life applications with Gorgias-B tool developed via the SoDA methodology. Explore seller agent policy examples, SoDA methodology considerations, and the methodological process involved in SoDA development.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
GorgiasB Argumentation in Practice Nikolaos Spanoudakis , Technical University of Crete, nikos@amcl.tuc.gr Antonis Kakas, University of Cyprus antonis@ucy.ac.cy Pavlos Moraitis, LIPADE, Paris Descartes University pavlos@mi.parisdescartes.fr
Contents Argumentation in Practice - Real-life Applications SoDA: Software Development for Argumentation Demo of Gorgias-B to support SoDA Seller Agent Preference-based argumentation framework of Gorgias Demo of Call Assistant in Gorgias-B: Cognitive Assistants.
Motivation Argumentation has matured for real-life applications Argumentation for Cognitive Systems Argumentation can benefit from a Systematic Software Development Methodology Real-life applications of Gorgias informs the SoDA methodology Gorgias-B is developed as a tool to enable applications of argumentation via the SoDA methodology.
Example application policy A seller agent policy is given as follows: Normally, sell a product at its high price. You can sell a product at the lower price only if payment is cash. Normally prefer to sell high. Regular customers can be offered the low price. Also the low price can be offered if the buyer can have late delivery. In high season you must sell at high prices. How can we capture this in an argumentation theory?
Seller Agent Gorgias-B demo
SoDA methodology In SoDA we need to consider the following questions: What is the decision problem? What are the options? What should we use as the object level arguments? What are the possible scenarios given the object-level argument rules? What are the contexts that refine the scenarios? Is the model/representation complete? How do we extend the model? With new refinement contexts (in existing scenarios) With new scenarios.
SoDA methodological process In SoDA the developer follows threads: 1. 2. Considers an application scenario States expected (preferred) output/decision Either unconditionally 2. Or in a Contextthat further refines the scenario Returns to (1) to consider the refined scenario(s) 1. 3. Termination of a thread: At 2.1 with thread end-open When refined joint scenario is inconsistent with end-final Threads can fork at point 2.1 Different expected decisions under different contexts
The thread generation process CHOOSE A PAIR OF OPTIONS: Oi, Oj (from WP1) START YES Cij= Cji= CHOOSE SET OF CONDITIONS Ci, Cjs.t. Choose Oiwhen Ci, choose Ojwhen Cj S1= Ci Cjis consistent (from WP4) NO END-THREAD-OPEN K=K+1 END-THREAD-FINAL SK =SK-1 CK-1ij CK-1ji K=1 NO CHOOSE CONDITIONS Cij, Cjis.t. Oi Ojwhen SK CKijis consistent Oj Oiwhen SK CKjiis consistent (from WP3, level K) SK consistent YES
Cognitive Assistants Call Assistant Demo A call assistant behaviour policy is given as follows: In general all calls are accepted. When at work I do not want to take family calls. However, if they are from my son I do want them but not when I am in a meeting. If he is at school and I believe that he is ill I want to take the call even if I am in a meeting. I will not take a call when I am in a family outing and the call is related to business. Unless it is the case that my boss is calling, when I will take the call. See Gorgias-B website.
Cognitive Call Assistant Demo Scenarios and Contexts in levels: Call: Family or Business State: At work or With Family Call From: Son; State: Son at school; Call From: Boss State: Family Outing Call From: Son; State: Son at school; In meeting; Call From: Son; State: Son ill;
Cognitive Assistants Call Assistant Demo
Seller Agent Gorgias Code complement(sell(Prd, Ag, high), sell(Prd, Ag, low)). complement(sell(Prd, Ag, low), sell(Prd, Ag, high)). rule(r1(Prd, Ag, high), sell(Prd, Ag, high), []). rule(r2(Prd, Ag, low), sell(Prd, Ag, low), [pay_cash(Ag, Prd)]). rule(p1_r1_r2(high, Prd, Ag), prefer(r1(Prd, Ag, high), r2(Prd, Ag, low)), []). rule(p2_r2_r1(Prd, low, Ag), prefer(r2(Prd, Ag, low), r1(Prd, Ag, high)), []):- regular(Ag). rule(c1_p1_r1_r2_p2_r2_r1(high, Prd, Ag), prefer(p1_r1_r2(high, Prd, Ag), p2_r2_r1(Prd, low, Ag)), [high_season]). rule(c2_p2_r2_r1_p1_r1_r2(Prd, low, Ag), prefer(p2_r2_r1(Prd, low, Ag), p1_r1_r2(high, Prd, Ag)), [neg(high_season)]). abducible(pay_cash(_, Prd), []). abducible(neg(pay_cash(_, Prd)), []).
Algorithm for generating Argumentation Gorgias Code
Argumentation in Gorgias Argumentation Framework of Gorgias Rules and Priorities rules (on rules) Preference-based argumentation framework Logic Programming with Negation as Failure (LPwNF) Proposed in 1994 (KMD at ICLP94) LPwNF used in the study of various problems Autonomous Agents (Computeesin SOCs) Machine Learning (Non-monotonic learning) Reasoning about Actions and Change (Event Calculus and Language E)
Argumentation framework of LPwNF Language of LPwNF Rules and Priorities on rules Conditional priorities Higher-order priorities, i.e. priorities on priorities C.f. Prakken & Sator, 1997: Legal Reasoning & Argumentation Uniform argumentation process for object-level and priority arguments Start argumentation at the object-level to support a decision Continue arguing at the meta-level on the relative strength of the arguments at the previous level. Uniform inclusion of abductive reasoning within the AF
Argumentation framework of LPwNF Arguments are tuples (or sets) S=(O,P) O object level Extended LP rules (without NAF) P priority rules on rules of O and P Attacking relation: S attacks S iff there exist L and S1 S, S 1 S s.t.: B S1 minL and B S 1 minLc S1 S 1 (If S1has a rule of lower priority then it also has one of higher priority) Either S renders a rule in S 1higher priority over a rule in S1 Or the rules in S 1and S1are non-comparable w.r.t. S and S
Temporal Reasoning (Reasoning about Actions) Arguments in A are: generation rules: holds (F,T2) initiation(F,T1), T1< T2 holds (F,T2) termination (F,T1), T1< T2 persistence rules: holds (F,T2) holds (F,T1), T1< T2 holds (F,T2) holds (F,T1), T1< T2 assumption rules: holds (F,T ), holds (F,T ) Priority < is given by: persistence rules < later(or =) & conflicting generation rules generation rules < later & conflicting generation rules assumptions < conflicting generation rules
Example Vaccinations Demo Gorgias InjectA initiates Protected when {TypeO} InjectB initiates Protected when { TypeO} InjectA happens-at 2 InjectB happens-at 3
Applications of Gorgias (2003- ) Ambient Intelligence: Ambient Assisted Living (AAL) Ambient Intelligence: Pervasive Services and Conflict resolution in sensors Business Computing: Product Pricing Business Computing: Portfolio Construction Network Security: Management of Firewall Policies Medical Informatics: Deep Vein Thrombosis PROSOCS platform for KGP agent: Intra-agent control
Find out more http://gorgiasb.tuc.gr/
Conclusions Cognitive expression of expected result(s) in an ever more specific problem scenarios corresponds naturally to an argumentation theory (in the LPwNF LPP framework). Facilitates applications of argumentation. Gorgias-B (www.gorgiasb.tuc.gr) is a tool to support this.
Cognitive Assistants via Argumentation Cognitive Assistant Architecture: Strategy Module World Knowledge Module (Beliefs and assumptions) Strategy Module: seen this in the demos. World Knowledge Module: Similar argumentation based representation: Common sense knowledge (Including Reasoning about Actions) Expert knowledge (in high-level cognitive terms) Supported by SoDA (and Gorgias-B) in the same way. MINED FROM THE WEB FROM UNSTRUCTURED (text) DATA.
Future Work Improve system of Gorgias Web-service New applications (with Imperial College Resilient Information Systems Security group) Security: Distributed Firewalls and Data Access Data sharing agreements Cyber attack attribution Cognitive Programming Cognitive Assistant applications Based on Comprehension through Argumentation STAR system (http://cognition.ouc.ac.cy/star)
